InclusionAI / inclusionai/ling-3.0-flash-fin

Ling 3.0 Flash Fin - access through LLMTR

Ling 3.0 Flash Fin is InclusionAI's finance-tuned build of Ling 3.0 Flash, activating roughly 5.1 billion of its 124 billion parameters per token. It is tuned for multi-step investment research and long-horizon planning and execution while keeping the base model's reasoning, coding and mathematics ability. It reasons by default, returning that chain separately in `reasoning_content`, whose tokens are counted inside `completion_tokens`. To turn reasoning off, send `reasoning_effort: "none"` or append the `:none` suffix to the model id - that is the only level with a measured effect on this model, so the others are not offered even though the upstream accepts them. It supports native tool calling and reads repeated prefixes from a prompt cache; note that with reasoning off the model does not honour `required` or named function choices in `tool_choice`, so that combination is refused with a 400 - keep reasoning on when you need to force a tool call, or use `auto` when you need to run with reasoning off. A single response can be at most 32,768 tokens. It accepts text only - no image or audio input, and no schema-enforced JSON output.